Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

Working on a tool to generate low poly collision meshes from gaussian splats. It uses marching cubes and quadric error metric (QEM) decimation. Results are pretty good, it can get a usable sub 10Mb collision mesh from a 50Mb splat.

martin_casado's profile picture

martin_casado

80,303 subscribers

49,759 görüntüleme • 6 ay önce •via X (Twitter)

Bilim & Teknoloji
LIVE

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

The low poly meshes I created are much, much faster with physics / collision than those directly from the splat to mesh translation. Amazing. I now have an end to end workflow for generated 3D worlds with usable collisions. Works great in VR and mobile too ...
0:36

The low poly meshes I created are much, much faster with physics / collision than those directly from the splat to mesh translation. Amazing. I now have an end to end workflow for generated 3D worlds with usable collisions. Works great in VR and mobile too ...

martin_casado

54,627 görüntüleme • 6 ay önce

The PlayCanvas team has solved collision for 3D Gaussian splats. Install splat-transform via NPM to get a CLI tool + library that can output high quality voxel-based collision. Here you can see a splat navigated in first person mode with voxel rendering toggled on/off. 🧵
0:37

The PlayCanvas team has solved collision for 3D Gaussian splats. Install splat-transform via NPM to get a CLI tool + library that can output high quality voxel-based collision. Here you can see a splat navigated in first person mode with voxel rendering toggled on/off. 🧵

Will Eastcott

30,263 görüntüleme • 3 ay önce

Collisions are finally coming! Automatic splat to conversion is tough given how messy AI generated splats are. So I created a simple editor that does down sampling, opacity filtering and then basic marching cubes and the results are pretty OK! 🥳
0:57

Collisions are finally coming! Automatic splat to conversion is tough given how messy AI generated splats are. So I created a simple editor that does down sampling, opacity filtering and then basic marching cubes and the results are pretty OK! 🥳

martin_casado

46,726 görüntüleme • 7 ay önce

splat-transform is a free and open source library and CLI tool for processing 3D Gaussian splats. It can create lightweight, accurate, voxel-based collision at scale. No human intervention required. This gym scene's voxels are perfectly generated - even large fully transparent windows are correctly filled. The voxel data file is 118KB. Start building 3D Gaussian splat apps today with PlayCanvas and splat-transform. 🧑‍💻
0:43

splat-transform is a free and open source library and CLI tool for processing 3D Gaussian splats. It can create lightweight, accurate, voxel-based collision at scale. No human intervention required. This gym scene's voxels are perfectly generated - even large fully transparent windows are correctly filled. The voxel data file is 118KB. Start building 3D Gaussian splat apps today with PlayCanvas and splat-transform. 🧑‍💻

Will Eastcott

20,508 görüntüleme • 2 ay önce

3D Gaussian Splats are cool, but they're static (Part 25). DG-Mesh can reconstruct high-quality and time-consistent 3D meshes from a single video and track mesh vertices over time, which enables texture editing on dynamic objects.
0:22

3D Gaussian Splats are cool, but they're static (Part 25). DG-Mesh can reconstruct high-quality and time-consistent 3D meshes from a single video and track mesh vertices over time, which enables texture editing on dynamic objects.

Dreaming Tulpa 🥓👑

79,033 görüntüleme • 2 yıl önce

Voxel collision in SuperSplat is a game changer. 🧊 It turns any Gaussian splat into accurate 5 cm voxels, automatically. It's super-compressed and far more efficient than a mesh, yet solid enough to climb stairs and walk crossbeams. Free & open source from PlayCanvas 👇
0:42

Voxel collision in SuperSplat is a game changer. 🧊 It turns any Gaussian splat into accurate 5 cm voxels, automatically. It's super-compressed and far more efficient than a mesh, yet solid enough to climb stairs and walk crossbeams. Free & open source from PlayCanvas 👇

Will Eastcott

90,777 görüntüleme • 1 ay önce

Sparkjs streaming+LoD is now stable(ish). This flythrough is over 100m splats rendered in browser. The splats are chunked into ~900 pages with a working set of around 16million splats. I think it can handle arbitrary size worlds now. Anyone have a 1 billion splat I can try?
1:31

Sparkjs streaming+LoD is now stable(ish). This flythrough is over 100m splats rendered in browser. The splats are chunked into ~900 pages with a working set of around 16million splats. I think it can handle arbitrary size worlds now. Anyone have a 1 billion splat I can try?

martin_casado

27,107 görüntüleme • 7 ay önce

Wild. Single shot text to 3D scene (gaussian splat) with no editing. This is approaching a usable scene asset for a game / 3D application.
0:21

Wild. Single shot text to 3D scene (gaussian splat) with no editing. This is approaching a usable scene asset for a game / 3D application.

martin_casado

56,136 görüntüleme • 1 yıl önce

What happens if you try to push and squash a Gaussian splat? :) Been playing with adding simplified deformation and a lightweight physics simulation layer to Gaussian splats, running directly in real time on mobile. Instead of simulating every splat, this uses a small CPU lattice to drive the deformation, plus a per-splat transform so they actually stretch and compress properly. No mesh conversion, no retraining. Just layered on top of the renderer. You can push, bend, squash a scan and it holds together surprisingly well... Still early, but I can see the potential in making splats feel more real, from static captures to something more tactile, something you can touch, throw, bounce around… or well just fold and squash on the floor :) #GaussianSplatting #3DGS #Unity
0:39

What happens if you try to push and squash a Gaussian splat? :) Been playing with adding simplified deformation and a lightweight physics simulation layer to Gaussian splats, running directly in real time on mobile. Instead of simulating every splat, this uses a small CPU lattice to drive the deformation, plus a per-splat transform so they actually stretch and compress properly. No mesh conversion, no retraining. Just layered on top of the renderer. You can push, bend, squash a scan and it holds together surprisingly well... Still early, but I can see the potential in making splats feel more real, from static captures to something more tactile, something you can touch, throw, bounce around… or well just fold and squash on the floor :) #GaussianSplatting #3DGS #Unity

𝐑𝐔𝐁𝐄𝐍🥽𝐅𝐑𝐎

51,284 görüntüleme • 4 ay önce

Instant Skinned Gaussian Avatars for Web, Mobile and VR Applications Short summary: In our system, we animate a background 3D mesh and have the Gaussian splats follow the mesh’s vertices. During preprocessing, splats are assigned to mesh vertices, and their relative transformations are stored. Once this data is saved, you can instantly use it in your applications without further preprocessing. At runtime, we animate the background 3D mesh, update the Gaussian splats in parallel, and resort all Gaussian splats every frame based on the viewer’s perspective.
1:10

Instant Skinned Gaussian Avatars for Web, Mobile and VR Applications Short summary: In our system, we animate a background 3D mesh and have the Gaussian splats follow the mesh’s vertices. During preprocessing, splats are assigned to mesh vertices, and their relative transformations are stored. Once this data is saved, you can instantly use it in your applications without further preprocessing. At runtime, we animate the background 3D mesh, update the Gaussian splats in parallel, and resort all Gaussian splats every frame based on the viewer’s perspective.

MrNeRF

66,016 görüntüleme • 10 ay önce

AI aided 3D world building is getting ridiculous. This is a large, multi-room gaussian splat (~6million splats). Has streaming, LoD, character support, spatial audio, physics, collisions, VR support, and streaming video. Audio on ... Great place to watch a movie :)
0:58

AI aided 3D world building is getting ridiculous. This is a large, multi-room gaussian splat (~6million splats). Has streaming, LoD, character support, spatial audio, physics, collisions, VR support, and streaming video. Audio on ... Great place to watch a movie :)

martin_casado

47,602 görüntüleme • 7 ay önce

AI 3D workflows for web are getting scary good. Here is a demo with a large splat background (+LoD), mesh character, spatial audio, VR support, and lighting. I was able to create it in a few hours (repo in reply). built with Three.js spark World Labs Audio on :)
1:03

AI 3D workflows for web are getting scary good. Here is a demo with a large splat background (+LoD), mesh character, spatial audio, VR support, and lighting. I was able to create it in a few hours (repo in reply). built with Three.js spark World Labs Audio on :)

martin_casado

82,670 görüntüleme • 8 ay önce

SuperSplat's voxel collision is a game changer for 3D Gaussian splat creators. Generate a highly accurate, inescapable voxel representation of any splat, regardless of quality. Generation is performed on the GPU, so it's super-fast. The voxel representation is tiny - only 418KB for this scene. And all this magic is free and open source, courtesy of the splat-transform library from PlayCanvas.
0:50

SuperSplat's voxel collision is a game changer for 3D Gaussian splat creators. Generate a highly accurate, inescapable voxel representation of any splat, regardless of quality. Generation is performed on the GPU, so it's super-fast. The voxel representation is tiny - only 418KB for this scene. And all this magic is free and open source, courtesy of the splat-transform library from PlayCanvas.

Will Eastcott

12,653 görüntüleme • 3 ay önce

I built a new mesh gradient tool that can create beautiful gradients from photographs. The results are often dreamlike and capture a general vibe. I'm calling it Photo Gradient, it's free on web and soon a Figma plugin, link below
0:58

I built a new mesh gradient tool that can create beautiful gradients from photographs. The results are often dreamlike and capture a general vibe. I'm calling it Photo Gradient, it's free on web and soon a Figma plugin, link below

Daniel Destefanis

103,132 görüntüleme • 2 yıl önce

Well, it appears that turning a Gaussian Splat into a #3D mesh was not that hard after all 😅. I have recently tried an app called KIRI Engine - 3D Scanner App and it offers several nice features that for me made it stand out from many other scanning apps: - It supports both photos and videos - You can clean up the splat directly in the app - It works on both Android, iOS and web! - BUT MOST IMPORTANTLY you can export the mesh making gaussian splitting a much more usable format for various production pipelines. I will be interviewing Jack Wang, the CEO of Kiri for an upcoming episode of XR AI Spotlight: What questions should I ask him?
0:59

Well, it appears that turning a Gaussian Splat into a #3D mesh was not that hard after all 😅. I have recently tried an app called KIRI Engine - 3D Scanner App and it offers several nice features that for me made it stand out from many other scanning apps: - It supports both photos and videos - You can clean up the splat directly in the app - It works on both Android, iOS and web! - BUT MOST IMPORTANTLY you can export the mesh making gaussian splitting a much more usable format for various production pipelines. I will be interviewing Jack Wang, the CEO of Kiri for an upcoming episode of XR AI Spotlight: What questions should I ask him?

Gabriele Romagnoli

22,997 görüntüleme • 1 yıl önce

New in Notch 2026.2: fully interactive Gaussian Splats. Imported objects and environments can be lit, reflected, and cast shadows, sitting naturally alongside any other element in the production. In this sample, we used a splat of a library cafe with a few 3D objects dropped inside to show it in action. Grab the sample scene from the Launcher and try it for yourself. Splat by:
0:20

New in Notch 2026.2: fully interactive Gaussian Splats. Imported objects and environments can be lit, reflected, and cast shadows, sitting naturally alongside any other element in the production. In this sample, we used a splat of a library cafe with a few 3D objects dropped inside to show it in action. Grab the sample scene from the Launcher and try it for yourself. Splat by:

Notch

11,280 görüntüleme • 1 ay önce

Some of the large scene's being generated are really unbelievable. This is a generative 3D splat, not 2D video. The full scene is likely over 50million splats. You can see some stitching inconsistencies from the final v2v pass. Also nice music :)
1:06

Some of the large scene's being generated are really unbelievable. This is a generative 3D splat, not 2D video. The full scene is likely over 50million splats. You can see some stitching inconsistencies from the final v2v pass. Also nice music :)

martin_casado

138,482 görüntüleme • 11 ay önce

Multi-splat object LOD test in sparkjs. 7 AI generated worlds, 12.5m total splats. Pretty consistent 120 fps on my laptop.
0:52

Multi-splat object LOD test in sparkjs. 7 AI generated worlds, 12.5m total splats. Pretty consistent 120 fps on my laptop.

martin_casado

95,404 görüntüleme • 9 ay önce

More and more users are turning their 360 footage into fully explorable 3D Gaussian Splats with View it here: This reconstruction started from a simple capture and is now a fully explorable 3D Gaussian Splat rendered in the FreeGaussian viewer, complete with point cloud reconstruction and high-fidelity spatial detail. More and more creators are discovering that their existing footage can become immersive 3D scenes. And they’re doing it for free with FreeGaussian
0:58

More and more users are turning their 360 footage into fully explorable 3D Gaussian Splats with View it here: This reconstruction started from a simple capture and is now a fully explorable 3D Gaussian Splat rendered in the FreeGaussian viewer, complete with point cloud reconstruction and high-fidelity spatial detail. More and more creators are discovering that their existing footage can become immersive 3D scenes. And they’re doing it for free with FreeGaussian

Over the Reality 🌐

14,759 görüntüleme • 3 ay önce

Flexible Isosurface Extraction for Gradient-Based Mesh Optimization paper page: This work considers gradient-based mesh optimization, where we iteratively optimize for a 3D surface mesh by representing it as the isosurface of a scalar field, an increasingly common paradigm in applications including photogrammetry, generative modeling, and inverse physics. Existing implementations adapt classic isosurface extraction algorithms like Marching Cubes or Dual Contouring; these techniques were designed to extract meshes from fixed, known fields, and in the optimization setting they lack the degrees of freedom to represent high-quality feature-preserving meshes, or suffer from numerical instabilities. We introduce FlexiCubes, an isosurface representation specifically designed for optimizing an unknown mesh with respect to geometric, visual, or even physical objectives. Our main insight is to introduce additional carefully-chosen parameters into the representation, which allow local flexible adjustments to the extracted mesh geometry and connectivity. These parameters are updated along with the underlying scalar field via automatic differentiation when optimizing for a downstream task. We base our extraction scheme on Dual Marching Cubes for improved topological properties, and present extensions to optionally generate tetrahedral and hierarchically-adaptive meshes. Extensive experiments validate FlexiCubes on both synthetic benchmarks and real-world applications, showing that it offers significant improvements in mesh quality and geometric fidelity.
0:51

Flexible Isosurface Extraction for Gradient-Based Mesh Optimization paper page: This work considers gradient-based mesh optimization, where we iteratively optimize for a 3D surface mesh by representing it as the isosurface of a scalar field, an increasingly common paradigm in applications including photogrammetry, generative modeling, and inverse physics. Existing implementations adapt classic isosurface extraction algorithms like Marching Cubes or Dual Contouring; these techniques were designed to extract meshes from fixed, known fields, and in the optimization setting they lack the degrees of freedom to represent high-quality feature-preserving meshes, or suffer from numerical instabilities. We introduce FlexiCubes, an isosurface representation specifically designed for optimizing an unknown mesh with respect to geometric, visual, or even physical objectives. Our main insight is to introduce additional carefully-chosen parameters into the representation, which allow local flexible adjustments to the extracted mesh geometry and connectivity. These parameters are updated along with the underlying scalar field via automatic differentiation when optimizing for a downstream task. We base our extraction scheme on Dual Marching Cubes for improved topological properties, and present extensions to optionally generate tetrahedral and hierarchically-adaptive meshes. Extensive experiments validate FlexiCubes on both synthetic benchmarks and real-world applications, showing that it offers significant improvements in mesh quality and geometric fidelity.

AK

38,450 görüntüleme • 3 yıl önce